Supporting information for: Real-Time Tunable Colors from Microfluidic Reconfigurable All-Dielectric Metasurfaces

Transcription

1 Supporting information for: Real-Time Tunable Colors from Microfluidic Reconfigurable All-Dielectric Metasurfaces Shang Sun #, Wenhong Yang #, Chen Zhang, Jixiang Jing, Yisheng Gao, Xiaoyi Yu, Qinghai Song*, and Shumin Xiao* State Key Laboratory on Tunable Laser Technology, Ministry of Industry and Information Technology Key Lab of Micro-Nano Optoelectronic Information System, Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen, , China, Collaborative Innovation Center of Extreme Optics, Shanxi University, Taiyuan, , China. * shumin.xiao@hitsz.edu.cn; * qinghai.song@hitsz.edu.cn Part 1. Refractive index of TiO 2 High quality TiO 2 films were essential for this research. In our experiment, the TiO 2 film was formed by electron-beam (E-beam) evaporation. The deposition rate was 0.8 Å/s and the base vacuum pressure was kept at torr. The optical constants of deposited TiO 2 were determined from spectrometry ellipsometry measurements. The data was collected at angles of incidence of 50 degree, 60 degree, and 70 degree, respectively. By fitting the measured spectra with calculated spectra using a multilayer model, the refractive index at different wavelength was obtained by modeling with Cauchy dispersion equations. After fitting, the calculated spectra were perfectly matched with the measured ones at all wavelength and all angles. Thus the optical constants and thickness of the E-beam evaporated TiO 2 film can be finally determined. Figure S1 shows the refractive index and extinction coefficient of TiO 2 film. The refractive index (n) of the TiO 2 film varies from 2.45 at 400 nm to 2.15 at 800 nm. Meanwhile, the extinction coefficient (k) remains almost zero in the whole visible spectral range. The relatively large refractive index and low extinction coefficient ensure the applications of deposited TiO 2 in all-dielectric metasurfaces.

2 Figure S1. Optical constants of TiO 2. The refractive index (n) and extinction coefficient (k) is illustrated as black solid line and red dashed line, respectively. Part 2. Numerical simulations and discussions Commercially available finite element method (FEM)-based software, COMSOL Multiphysics, was used to realize all the numerical calculations in this work. Periodic boundary conditions were used in the software to simulate the periodic array of the metasurface with one unit cell. TiO 2 antenna array in our work was settled on top of 15-nm ITO glass substrate, with the optical constant of TiO 2 determined from spectroscopic ellipsometry measurement. The reflection spectra of the metasurface were taken with the background refractive index varying from 1 to , and , representing the circumstance of air, water, dimethyl sulfoxide and carbon disulfide. In order to match the experiment results, the recorded color (R*Lightsource) would be the combination of the spectrum of the lamp (Lightsource, shown as the dash line in Figure S2) and the reflectance spectrum of the structure R. Based on the spectra, colors can be calculated in the CIE 1931 chromaticity map. We take the sample with period equal to 350 nm for instance, both the spectra and the colors of R and R*Lightsource are illustrated. Figure S2. Normalized reflection spectrum (R) and the corrected reflection spectrum with halogen light-source (R*Lightsource). The orange dashed line shows the spectrum of the halogen lamp in the experiments. The insets show the corresponding calculated colors.

3 In our design, the ITO layer won t help the light confinement due to its relatively high refractive index (n ~ 2.0 at 400 nm and ~ 1.8 at 650 nm). This is more clearly to see in numerical calculations. We have numerically calculated the reflection spectra of a TiO 2 metasurface with p=300 nm, w=230 nm for instance. All the results are summarized in Figure S3. Without the ITO layer, the reflection peak is nearly 100%. Once 15 nm ITO layer is added, the peak value reduced to ~ 95% and the full width at half maximum becomes broader. The performance is further reduced when the thickness of ITO layer is increased. All these results show that the presence of ITO layer spoil the performances of the nanostructures. However, from the point view of experiment, the ITO is quite important to conduct the electron during the electron beam lithography. Without the ITO layer, the glass is not conductive and it is hard to realize nanosized structures and the backscattering of electrons will also be too strong to form any nanostructures. In our experiment, we have to trade off the spoil of reflection peak and the fabrication tolerance. Therefore, a relatively thin ITO layer has been employed experimentally. Figure S3. The numerically calculated reflection spectrum from TiO 2 metasurface on a glass substrate without (a) and with 15 nm ITO film. Here the period p, nanoblock width w, and the height are 300 nm, 230 nm, and 250 nm, respectively. Additional field profiles of E x for electric dipole and H y for magnetic dipole at the x-y cross-sections are shown in Figure S4. When the nanostructure is illuminated with an x polarized light in normal direction, both of the field profiles in x-y cross-section (see the arrows) and x-z cross-section show that the electric fields are dominated by E x. The magnetic field is circulating around x-axis. Similarly, both of the magnetic field distributions in x-y cross-section and x-z cross-section (see the arrows) show that the magnetic field goes along y direction and the electric field goes circling around y axis. All these results demonstrate the properties of ED and MD very well.

4 Figure S4. The electric and magnetic fields of the cross-section at the ED and MD resonances respectively. In case of tunable color, we note that TiO 2 metasurfaces have their own advantages in dynamic colors. Compared with silicon, TiO 2 has relatively lower refractive index and thus more electromagnetic fields are distributed in the gap regions. As a result, the responses of TiO 2 metasurfaces should be more significant than the Si metasurfaces. We have numerically confirmed this information. Here the structural information are taken from Proust et al. report (ACS Nano 2016, 10, ) and two metasurfaces are designed to have similar resonant wavelengths. As shown in Figure S5 below, with the increase of environmental refractive index from 1 to 1.6, the peak wavelength of Silicon shifts 28 nm, which is much smaller than its full width at half maximum (FWHM). For a direct comparison, the reflection peak of TiO 2 metasurfaces with similar structural color shifts 56 nm, which is much larger than the FWHM and can be simply resolved in displayed colors. Thus due to the lower refractive index of TiO 2, the corresponding color transition is more dramatic. Figure S5: The wavelength shift of Si metasurface (a) and TiO 2 metasurface (b) by changing the environmental refractive index from 1.0 to 1.6.

5 In Figure 5, two kinds of metasurfaces with different periods p = 380 nm and p = 400 nm were fabricated. The reflection spectra for these two metasurfaces have been numerical simulated, and summarized in Figure S6. Associated with the wavelength shift, we have also included the CIE color matching functions in Figure S6. When the metasurfaces are surrounded by air, the metasurface with p=380 nm has almost equal contributions from green and red colors, resulting in an orange color. One the contrary, the metasurface with p=400 nm is dominated by the red color. Thus the color contrast is distinct and can be clearly seen. With the increase of environment refractive index, the peaks for two metasurfaces move into the gray area, especially for the CS 2 case. In that region, the ratio of green and red colors are very close in two metasurfaces, leading to two similar orange colors (see the insets). Thus the encoded information is concealed. Figure S6. Simulation reflection spectra for the two TiO 2 metasurfaces in Figure 5 within different solvents and color matching functions. The insets are the corresponding structural colors. Furthermore, we have numerically calculated the resonant properties of TiO 2 metasurfaces with different period numbers. Here the lattice constant and the size of nanoblock are p=300 nm, w=230 nm. All the results are summarized in Figure S7 below. With the decrease of the array number, the reflection peak around 475 nm becomes weaker and weaker, with the resonant wavelength unchanged. This information makes the underlying mechanism of our design very clear. The reflection peak at 475 nm is mainly caused by the coupling between magnetic resonance (electric resonance) with radiation from periodic nanostructure. At smaller period, while the magnetic resonance and electric resonance from TiO 2 nanoparticle keep the same, the reflection from periodic structure becomes weaker.

6 Consequently, the reflection peak in Figure S7 reduces quickly. Fortunately, the structural colors don t require extremely high reflection and can still be formed with small period such as 4 4. Figure S7. Simulated reflection spectra for different periodicity. Part 3. The fabrication of TiO 2 metasurfaces The metasurfaces were fabricated with electron beam lithography technique followed by lift-off process. First, we cleaned the 15nm-ITO glass substrates in ultrasound bath in acetone and isopropyl alcohol (IPA) for 10min respectively and dried under clean nitrogen flow. Then, 350 nm ZEP520 film was spin-coated onto the ITO-coated glass substrate and baked at 180 for an hour. After that the sample is exposed to electron beam in E-beam writer (Raith E-line, 30 kv) and developed in ND510 solution for 60 seconds at 0 to form the ZEP nanostructures. Then the sample was transferred into an E-beam evaporator and directly coated with 200 nm TiO 2 films (deposition rate 0.8 Å/s, base vacuum pressure torr). After immersing the sample in acetone for 8 hours, the ZEP was removed and the reversed nanostructures were well transferred to TiO 2 film. Exposed to oxygen plasma for 5 min to remove the remaining ZEP around nanostructures, TiO 2 metasurfaces were finally formed. Finally the samples were transferred into a tubular furnace, where a thermal annealing at 300 was performed for 1 hour under oxygen atmosphere to make the surface of the TiO 2 nanostructure smoother and decrease the radiation loss from the surface roughness. Then, a PDMS microfluidic channel was fabricated and integrated with the sample. We chose the silicon wafer as the template of microfluidic channel. The silicon wafer is cleaned with oxygen plasma for 5 minutes to increase its hydrophilicity. A layer of photoresist (SU-8) about 500 µm is dropped onto the silicon wafer and baked at 96. After UV exposure for 80 seconds, the microfluidic channel mask is achieved after a 3 minutes post exposure baking (PEB) and a development for 30 minutes. The dimensions of the microfluidic channel are 10 mm 4 mm 500 µm. In order to ensure the transparency, the PDMS (the mass ratio of

7 prepolymer and curing agent is 10: 1) microfluidic channel is placed into a glove box at vacuum with 0.08MPa about 10 min to eliminate air bubbles. Then the PDMS is solidified covering on the block baked at 80 for 30 min and the microfluidic channel is finally formed. The PDMS channel is transferred onto the metasurface sample, using the UV-exposure for 2 min and plasma cleaner with oxygen for 5 min to bond the PDMS and the sample. During the bonding process, the room temperature vulcanizing (RTV) silicone rubber was coated at the interface between PDMS and ITO glass to seal the microfluidic channel completely. Part 4. Optical Characterization The sample was placed onto a three-dimensional translation stage under a home-made microscope. The characteristics of the reflection spectra were took under the normal incidence with a spectrometer. The bright-field microscopy images were taken under an optical microscope (ZEISS, Axio Scope AI) with Canon EOS 600D. Figure S8 shows the setup for optical reflection characterization. Figure S8: Schematic picture of the setup for optical characterization. Part 5. Tunable colors of TiO 2 metasurfaces The reflection spectra for samples in water, DMSO and CS 2 have been calculated and experimentally measured. The corresponding colors were calculated as well. All the results are illustrated in Figure S9-S11 for the responses in water, DMSO and CS 2 respectively. From the figures, the spectra from simulations and experiments match well, and we can see that the microfluidic reconfigurable structural colors can be realized in almost the entire visible spectrum.

8 Figure S9. (a) (b) Simulation and experiment reflections for samples in water. (c) The corresponding photograph for each sample. Figure S10. (a) (b) Simulation and experiment reflections for samples in DMSO. (c) The corresponding photograph for each sample.

9 Figure S11. (a) (b) Simulation and experiment reflections for samples in CS 2 (c) The corresponding photograph for each sample. Part 6. Influences from photoinduced hydrophilicity In our experiment, the hydrophilic surface of TiO 2 metasurface is quite important. We have numerically studied the influences of hydrophilic surfaces of TiO 2 metasurfaces. In case of hydrophobic surface, the gaps between TiO 2 nano-blocks cannot be fully filled with solution. To mimic this process, we use a parameter h to describe the fill fraction of water surrounding the structures (see the schematic picture in Figure S12). Figure S12. The schematic picture to define the filling fraction of water inside TiO 2 metasurface. The unfilled region is defined with a height parameter h.

10 We took a metasurface with p=350 nm, w=290 nm as an example to show the impacts of filling fraction. The other parameters are the same as Figure 3 in the main manuscript. The reflection spectra with h = 0 nm, 50 nm, 150 nm and 200 nm have been calculated and shown in Figure S13(a). We can see that both the peak positions and their relative amplitudes changed with the increase of h. For a direct comparison, we inserted the experimentally recorded reflection spectrum of the metasurface with p = 350 nm and w = 290 nm. We can see that both of the peak position and their relative amplitudes are very close to the simulation results with h=0 nm, indicating that our TiO 2 metasurfaces are fully covered by water in real experiment. The large filling fraction clearly show that the TiO 2 surfaces are hydrophilic. This deduction is also consistent with the experimental result in Figure S14(b), where the contact angle is only around 75 degree. The influences of UV exposure to the TiO 2 metasurface have also been studied by fabricating a new sample with p=350 nm, w=290 nm. We first took the spectrum of the sample in water. Then the sample was blow with N 2 gas. After that, the sample was exposed to a UV lamp at 365 nnm for 10 min. Here the UV exposure conditions followed reference Langmuir 2016, 32, The average value of radiation intensity reaching the surface measured with a radiometer was 8 mw/cm 2. Figure S13(b) demonstrated the reflection spectra of metasurface in water circumstance before and after UV exposure. No obvious difference was found, indicating that the UV illumination does not change the optical responses of our metasurfaces here. Figure S13. Simulation and experiment reflection spectra for the metasurface. (a) Simulation and experiment reflection spectra with different filling fraction. (b) The reflection spectra of metasurface in water circumstance before and after UV exposure.

11 To understand the results in Figure S13(b), we fabricated two pieces of 200 nm thick TiO 2 films on glass substrates with same condition. One was radiated under UV illumination and another was not. Then the water static contact angles (SCA) were measured for the two TiO 2 samples with 5 µl droplets. Each sample was measured three times to get a reliable average value. The results are summarized in Figure S14. The SCA for sample before UV exposure is 75±2, after UV exposure is 68±3. We can see that the hydrophilicity is not significantly changed, consistent with the reflection spectra in Figure S13(b). Figure S14. The water static contact angles for the samples before and after UV exposure. Part 7. More experimental results for reconfigurable images To test the generality of reconfigurable color images, we have fabricated six patterns with different colors and encoded information. All of the experimental results are summarized in Figure S15 by injecting water, DMSO and CS 2. Here we took TiO 2 metasurfaces with p = 300 nm, g= 60 nm (for background) and p = 310 nm, g = 60 nm (for encoded information) to realize the image H. Similarly, TiO 2 metasurfaces with p = 350 nm, g= 60 nm and p = 360 nm, g = 60 nm have been utilized to build the image I. TiO 2 metasurfaces with p = 400 nm, g= 60 nm and p = 380 nm, g = 60 nm are used for the image HIT. With the increase of environmental refractive index, we can see that the structural colors keep red-shifting. Most importantly, the color contrast between information and the background reduce and the encoded information almost totally disappear in CS 2.

12 Figure S15. Simulation and experiment results for six patterns in air, water, DMSO and CS 2 environment.the structural parameters for image H are p = 300 nm, g= 60 nm and p = 310 nm, g = 60 nm. The structural parameters for image I are p = 350 nm, g= 60 nm and p = 360 nm, g = 60 nm. The structural parameters for image HIT are p = 400 nm, g= 60 nm and p = 380 nm, g = 60 nm Part 8. Real-time color transitions in water and DMSO Here we take the snapshots in videos with 240 fps to illustrate the real time color transition in different solvents. Here the TiO 2 metasurfaces are exactly the same as the ones in Figure 7 in the main manuscript. We can see that the color transition time is always around 16.6 ms, which is limited by our video camera. Interestingly, since the same TiO 2 metasurface has been utilized in different solvents, we also know that the structural colors can be recovered by rejecting the solvents (see Figure S16(a) and Figure S17(a)). Figure S16. Video snapshots for sample in water at different timing. The structural parameters are the same as samples in Figure S15.

13 Figure S17. Video snapshots for sample in DMSO at different timing. The structural parameters are the same as samples in Figure S15. Part 9. The back and forth color transition and the perspective of advanced dynamic colors We have also fabricated a new sample to test the ability of the microfluidic metasurface to be switched back from liquid environment to air environment. The sample was composed of a TiO 2 metasurface with p=350 nm and g=60 nm. In the experiment, CS 2 solution was injected into the microfluidic channel, and a video was taken. During the injection, small bubbles were formed within the channel and passed the metasurface. Screenshots of the video are summarized in Figure S18, which shows the process of transition from CS 2 environment to air environment and back to CS 2 environment again. We can see that the color changed from gold to yellow within 62.5 ms. At around ms, the air bubble passed the metasurface, the color changed back to gold color within 22 ms. Similar back and forth color transitions can also be seen within another 120 ms. We note that the microfluidic channel is quite simple in this experiment. With the help of the mature microfluidic channel technology, we believe that back and forth color transitions with high frequency are very promising. Figure S18. The real-time back and forth structural color transition. By injecting a small bubble, the color can be switched from orange to green and back to orange within 150 ms. This process ensures the possibility of microfluidic metasurface in high frequency color transitions. We also can see that with the boundary passing by, the quality of the images get bad in Figure 7 in the main manuscript. That s because the infiltration of liquid has slightly changed the surrounding refractive index and made the image out of focus in CS 2. This influence can be eliminated by using PDMS microfluidic channel with smaller thickness. For a simple demonstration, the CS 2 was injected and the images were recorded. Figure S19 (a) and (b) show the snapshots of TiO 2 metasurfaces with and without CS 2 solution. Due to the inverse design, we can see that the color contrast is improved after the infiltration of solution into microfluidic channel. Moreover, we have reduced the height of microfluidic channel. As a result, the images with and without CS 2 are both well recorded on the focus.

14 Figure S19. The images of TiO 2 metasurfaces with (a) and without (b) CS 2 in the microfluidic channel. Both images are clearly shown and the encoded information can be seen more clearly after the infiltration of CS 2solution. And as we all know, the microfluidic channel is a very important modern technology, which has been widely applied in medical diagnosis, medicine delivery, and biosensing. By using the TiO 2 metasurface within microfluidic channel, the optical sensing technique can be changed from the measurement of spectra to the simple detection of color images. Then the microfluidic sensors can be more cost-effective. Meanwhile, the microfluidic technique has also been well developed in past decades. According to the review papers on Chem. Rev. 2013, 113, and Anal. Chem. 2013, 85, , we know that the microfluidic channel can be sub-100 nm. Meanwhile, as schematically depicted below in Figure S20 and Figure S21, microsized droplets can be formed within the channel and they can be redirected to different channels by applying external controls such as inhomogeneous magnetic fields. The narrow channel size and the precise control of droplets in propagation can ensure the pixel size resolution. Meanwhile, the size of droplet and the propagation speed can improve the switching time of dynamic colors. In this sense, the combination of microfluidic channel and TiO 2 metasurfaces can be an effective approach to realize dynamic colors, especially for the applications in biomedical sensing.

15 Figure S20. The schematic picture for the formation of micro-droplets within the microfluidic channel. (Taken from reference Rep. Prog. Phys. 2011, 75, ) Figure S21. One example of redirecting droplets in microfluidic channel with inhomogeneous magnetic field. (Taken from reference Lab Chip 2009, 9, ) Part 10. The color transition across the entire visible spectrum Since the limitations of field distributions within the gap areas, the wavelength shift in our experiment is limited to around 50 nm. This kind of wavelength shift can switch the color between adjacent color regions in CIE map, e.g. blue-cyan and green-yellow et al. In case of applications such as information encryption, anticounterfeit, and biomedical sensing, this kind of changes are good enough. For the applications of color displays, a tunable color across the entire visible is required. The recent literatures show that the sizes of nanostructures are not changed during the stretching. Since the stretching can be realized by MEMs or on flexible substrate, the mechanically generated stretching can be a good supplementary. Below, we show that entire visible color tuning can be realized with a combination of mechanical stretching and microfluidic channel. To verify this information, we simulated the properties of five different metasurfaces on rigid substrate. As the size of nanoparticle is well preserved, here we set the width of TiO 2 at 250 nm and increase the period from 300 nm to 320 nm, 340 nm, 350 nm and 360 nm. These kind of increases in period correspond to a substrate stretch of

16 0, 6.67%, 13.33%, 16.67%, and 20%, respectively. As shown in Figure S22(a) below, the reflection peak increases from ~ 480 nm to 570 nm and the corresponding displayed color changes from blue to green. Then the period is fixed at 360 nm and different solutions such as water, DMSO and CS 2 are injected into the channel. We can see that the displayed colors further shifts to longer wavelength. Following the numerical calculations, the samples have been experimentally fabricated and the color transitions have been recorded. As shown in Figure S22(b), while the reflection peaks are a little bit broader than the simulation, we can clearly see the wavelength shift and the corresponding color transitions. Due to the increased bandwidth of reflection peak, the experimentally recorded color after the infiltration of CS 2 is already red. Figure S23 summarizes the structural colors in CIE 1931 color space. We can clearly see that the tunable color can cover the entire visible spectrum. Figure S22. (a) Simulated reflection spectra for 5 metasurfaces with width kept at 250 nm and lattice size increased from 300 nm to 360 nm. Simulated reflection spectra for metasurface with the period fixed at 360 nm and different solutions such as water, DMSO and CS 2 injected are also showed. (b) Experimental results of the fabricated TiO 2 metasurfaces. All the structural parameters are the same as (a). The insets are the bright field photographs.

17 Figure S23.The structural colors in CIE 1931 color space. Beyond that, we also have done a simple experiment to demonstrate the ability for our metasurface to achieve color tuning across the RGB range. We first fabricated a sample with p=300 nm, w=230 nm on a silver substrate, following with PDMS covering on the sample. After eliminated air bubbles with 0.08MPa vacuum and solidified, the PDMS with metasurface inside was peeled off and formed a TiO 2 metasurfaces on flexible substrate as shown in Figure S24. Figure S24. Schematic picture of TiO 2 metasurface on PDMS substrate. Then we took pictures of the sample with different stretch ratio at x direction under reflection mode microscope, shown in Figure S25. It can be seen that, with flexible substrate deforming, the structural color of blue at original status red shifted, got cyan at 15% stretched, green at 30% stretched and finally red at 30% stretched with CS 2 injected. So with the assist of flexible substrate, we also can get color tuning effect across the RGB spectrum.

18 Figure S25. Pictures of the sample with different stretch ratio at x direction under reflection mode microscope. Part 11. Top view SEM images of color images In Figure S26, 11 SEM images for 11 samples in Figure 3 were included. Figure S26. SEM images for 11 samples in Figure 3. The scale bar is 500 nm Figure S27. SEM images for the sample in Figure 4. (Logo printed with permission from Harbin Institute of Technology. Copyright 2016 Harbin Institute of Technology.)

19 In Figure S28, SEM images for H and I in Figure 7 were included. Figure S28. SEM images for H and I in Figure 6.